1 Introduction

Additive manufacturing (AM) offers unparalleled design flexibility, the ability to produce complex geometries with enhanced strength-to-weight ratios and minimized material waste [1]. Among the various AM technologies, laser-based powder bed fusion of metals (PBF-LB/M) has gained considerable interest from a wide range of industries, including aerospace, automotive, and biomedical [2]. The layer-wise approach enables the creation of objects with intricate and lightweight structures that were previously unattainable using conventional production techniques, such as subtractive machining and casting [3]. This makes PBF-LB/M processes particularly suitable for manufacturing high-value parts, such as aerospace components and medical implants tailored to individual patients’ anatomies [4, 5]. In addition, PBF-LB/M allows using a wide range of metal powders as feedstock material, including difficult-to-cut alloys. These capabilities open to applications where specific material characteristics play a crucial role in ensuring optimal product performance [6].

However, despite the discussed advantages, the full potential of PBF-LB/M remains not completely exploitable due to several challenges and limitations still affecting the process [7]. Indeed, different types of defects might occur because of sub-optimal process parameters and feedstock material quality in terms of powder size distribution, shape, and chemical status [8, 9]. In particular, internal porosities, cracks, and contaminations can negatively impact the mechanical properties and reliability of fabricated parts [10]. Moreover, residual stresses generated during the build process can lead to warping and distortions, limiting the dimensional accuracy of final products [11]. To effectively improve the actual performance of PBF-LB/M and overcome the current limitations, an in-depth understanding of the underlying process dynamics and of the relation between process parameters and defects/inaccuracies is needed. PBF-LB/M is characterized by complex interactions between the laser beam and powder particles, and gaining insights into these interactions is fundamental for process optimization. Furthermore, improving product quality is a key driver for advancing the adoption of PBF-LB/M in advanced applications. In this context, in-process monitoring solutions and post-process part quality evaluations are strongly required to improve the knowledge of these aspects and to enable a more sustainable, first-time-right, and zero-defect production.

In-process monitoring of PBF-LB/M allows for acquiring different types of data and information depending on the type of sensors used, providing insights into the process when the fabrication is taking place [12]. While the gathered information can effectively contribute to improving the process understanding and quality, accurately determining the position of potential defect onset sites remains crucial. In particular, their position should be determined relatively to the actual building direction and plane of fabrication, as well as coater and gas flow directions. When using off-axis optical monitoring techniques, the perspective deformation is one first issue to be dealt with for the accurate determination of the position of potential defects. The solutions proposed so far in literature to perform perspective corrections were based either on the use of reference objects characterized by a number of features with known relative positions [13] or on the exploitation of intrinsic features (e.g., corners, holes) already present in the building plate [14]. However, both these solutions have limitations, as described in the following.

  • The former solution requires the following steps: (i) opening the machine to insert and position the reference object onto the building plate, (ii) acquiring images of the reference object using the selected monitoring technique, and (iii) re-opening the machine to remove the object. Off-axis monitoring techniques are often used in stand-alone setups exploiting the available machine windows, which in many cases are located on the main machine door. In such cases, the need for opening the door after capturing images of the reference object might require moving the optical device, invalidating the obtained parameters for perspective correction in case the device positioning is not well repeatable.

  • The main limit of the latter solution is instead related to the fact that the building plate is commonly re-generated after the process by removing a thin layer of material, for instance, by machining and/or grinding. These operations can alter the shape of built-in features used for the perspective correction, which then might require frequent re-calibration.

Another relevant issue for the accurate determination of potential defects position is that the observed anomalies in the acquired monitoring signals/data may or may not lead to the presence of real defects in the built part [15]. To increase the understanding of such correlation between anomalies and actual defects, X-ray computed tomography (CT) is often employed as a reference post-process measurement technique [16, 17]. X-ray CT is a powerful non-destructive method for advanced dimensional analyses of both internal and external geometries, features, and defects, without the need for disassembling or cross-sectioning the parts [18]. CT capabilities are particularly relevant within the additive manufacturing process chain, as they enable the inspection of intricate and/or internal structures and defects, resulting in a fundamental tool for evaluating part quality after manufacturing and pivoting the process development as well [19]. Despite the relevant advantages of CT, limitations concerning spatial resolution and accuracy still need to be completely overcome [20]. To effectively investigate micro-scale features and defects, high-resolution CT scans are indeed essential. Micro-focus CT scanners are available and increasingly used in industry to achieve high spatial resolutions, which are however indirectly correlated with the envelope dimensions of the scanned object [21]: the smaller the object, the higher the achievable geometrical magnification and spatial resolution. On the other side, good dimensional accuracy—which is needed to obtain quantitative reliable data—is still challenging to be properly determined and improved [22]. This is especially true for additively manufactured parts, which are typically characterized by complex shapes and surface topographies that are difficult to measure with more accurate reference measurement instruments (such as tactile coordinate measuring machines (CMMs)) needed to quantify and reduce the CT measurement errors by comparison [23]. Moreover, such complex shapes and surfaces hinder the alignment quality of CT reconstructed volumes with respect to the actual position of the produced parts during the building process.

In order to overcome the issues discussed above, this research paper presents an innovative solution based on the development of a new multi-function building plate for PBF-LB/M that can be used for improving both in-process monitoring and post-process CT measurements, as well as their comparability. So far, special building plates for PBF-LB/M were proposed to integrate sensors for in situ data acquisitions, such as polyamide coated fibers for strain measurements [24], ultrasonic sensors for stress monitoring and crack detection [25], and thermocouples attached underneath the build plate to monitor the heat flow [26]. Differently from other proposals, the plate conceived in this work followed a design-for-metrology approach [27] with the original contribution of enabling (i) accurate and on-machine perspective correction when using off-axis optical monitoring, (ii) high-resolution post-process CT scans, (iii) accurate alignment procedures between monitoring data, post-process CT data, and building properties (such as building direction and plane, as well as coater and gas flow directions), and (iv) accurate CT measurements and comparison with in-process gathered datasets. The end goal will be to exploit the improved measurements to enhance the process understanding, the product quality, and the industrial applications of PBF-LB/M. The new building plate is described in Section 2 in terms of design requirements and conceptualization, design of separable inserts, production, and assembly of inserts into the base plate. Possible uses of the platform are discussed in Section 3, where some application examples are reported. Specific in-depth studies based on the reported application examples will be addressed in the future through dedicated publications.

2 New building plate

This section describes the new plate, starting with the general requirements and concept that can be applied independently from the specific AM machine (Section 2.1). Additionally, the plate specifically designed and produced for the PBF-LB/M machine used in this work will be presented (Section 2.2 for design and Section 2.3 for production and assembly).

2.1 Requirements and plate concept

The main scope of the new plate is to enable more accurate in-process and post-process measurements to gather meaningful insights and effectively enhance the knowledge and quality of the process of interest. This is made possible by following a design-for-metrology approach and incorporating the use of removable inserts. In more detail, the following requirements have been considered: (i) the inserts must be mountable into their respective pockets in the plate base by ensuring repeatability of the assembly; (ii) the inserts must feature reference geometries to align the reconstructed tomographic volume with respect to the actual building plane and direction as well as to the coater and gas flow directions; (iii) in order to exploit the geometrical features for precise alignment, the inserts need to be CT scanned together with the parts produced on their top surface; (iv) their dimensions should allow for high-resolution tomographic scans (a necessary condition for analyzing micro-scale internal porosity, cracks, and surface features and defects), and (iv) the plate should also integrate markers for the perspective correction needed for in-process monitoring when conducted with off-axis optical systems, with no more need for intermediate machine openings to insert external reference objects for the correction. The alignment of CT reconstructed models to the main directions relevant for describing the process is crucial for performing dimensional and geometrical measurements with datum features and for accurately correlating post-process CT data with in-process monitoring data when employing X-ray computed tomography to validate monitoring techniques [16].

As an additional remark, the concept of a building plate with removable inserts proposed in this article is also relevant concerning the fast and efficient restoration of the plate itself after use. In fact, the plate base can be immediately reused after the process completion by replacing the utilized inserts with new or regenerated ones. Furthermore, the markers for perspective correction will not be subjected to restoration operations, hence remaining unaltered after the process, with no need for frequent re-calibration. After CT scanning, the inserts can be restored by separating the manufactured component and removing a thin layer of material from the fabrication plane to regenerate a flat and smooth surface. This procedure allows for avoiding the removal of a material layer from the entire plate, resulting in significant material and energy savings.

2.2 Plate and insert design

The PBF-LB/M machine used in this work is a Sisma MySINT 100 (Sisma SpA, Italy). The selected material is H13 tool steel. The original plate is a 99 mm diameter disk, while the design of the new plate is shown in Fig. 1. The base of the plate has the same dimensions as the original plate but features 12 pockets to accommodate the same number of inserts. These pockets are characterized by a first cylindrical section with a diameter of 12 mm and a height of 11 mm, ending with a 0.5 mm deep chamfer. The second section consists of a cylindrical hole with a diameter of 5 mm and a depth of 2.25 mm, followed by a truncated cone hole, which serves for the insertion of a fixing screw in the direction opposite to that of the insert. Adjacent to the first cylindrical section is a partial cylindrical hole with a radius of 1.5 mm, which will be used for the correct insertion and alignment of the insert using a precision pin (as better explained later in this section). A number of small cylindrical holes with a diameter of 3 mm and a depth of 4 mm are distributed over the plate surface to be used as markers for the perspective correction as mentioned in Section 2.1 and addressed in Section 3.1.

Fig. 1
figure 1

Technical drawing of the proposed new building plate. Dimensions are reported in millimeters

Full size image

Figure 2 illustrates three different designs for the removable inserts to be assembled into the plate base. They all consist of a cylindrical body with a diameter of 12 mm, ending with a 45° chamfer, i.e., the negative counterparts of the respective pockets in the plate described above. Each insert is characterized by the presence of a partial cylindrical hole on its side, which is needed for aligning in a repeatable way the insert in the plate. Specifically, a precision pin is inserted during the assembly into the cylindrical hole formed by the hole portion of the insert and the complementary portion on the respective pocket. The assembly is finalized by putting a screw into the inserts’ internal threaded hole.

Fig. 2
figure 2

Representation of three inserts to be assembled to the new plate, with different diameters of the regions to be CT scanned: 6 mm (a), 8 mm (b), and 12 mm (c). Technical sketch and main quotes for the 8 mm insert, which is taken as an example (d)

Full size image

Two insert types are additionally characterized by cylindrical protrusions with diameters of 6 mm (Fig. 2a) and 8 mm (Fig. 2b, d), respectively. Such protrusions are the regions that are meant to be included in the CT scans, along with the produced parts. For this reason, they have been laterally flattened to break the symmetry and aid the alignment of the reconstructed CT volumes. On the other hand, the third insert does not contain any protrusion since the region to be included in the CT scans is a top portion of the insert itself. All inserts are slightly taller than the hosting pockets, to have margin for repeated uses and recoveries of the inserts themselves after being used in PBF-LB/M processes. This aspect is important also from the sustainability point of view, as the inserts can be used multiple times. Figure 3 motivates the decision to design different inserts, as it schematically shows the maximum geometrical magnification that can be obtained with the different inserts when used on a cone-beam CT system; in particular, the figure refers to the magnifications achievable with the CT system used in this work: a Nikon Metrology MCT225 metrological CT system, characterized by micro-focus X-ray tube with minimum focal spot size equal to 3 μm, 16-bit 2000×2000 detector with pixel size of 0.2 mm, high-precision encoders for axes control, and cabinet temperature controlled at 20±0.5 °C. With this CT system, the insert with the 6 mm diameter protrusion can be scanned with a minimum voxel size of 3 μm, the insert with the 8 mm diameter protrusion with a minimum voxel size of 4 μm, and the 12 mm insert with a minimum voxel size of 6 μm. The possibility of producing inserts of different sizes, even different from those presented in this section, makes the proposed plate concept highly versatile. The overall dimensions of the inserts and pockets, as well as their relative position, can also be varied if necessary to adapt to specific needs. In this work, the pockets have been specifically arranged as shown in the top view of Fig. 1, thus implying a regular grid sample positioning. This choice has been driven by the interest in investigating the influence on the fabricated parts of both the recoating and the gas flow (whose directions in the employed PBF-LB/M machine with respect to the plate are reported in Fig. 1).

Fig. 3
figure 3

Schematic representation of the three inserts positioned at their minimum distance with respect to the X-ray source, to achieve the highest possible CT voxel resolution with the CT system used in this work; in particular, the resulting voxel size is equal to 3 μm for the 6 mm insert (a), 4 μm for the 8 mm insert (b), and 6 μm for the 12 mm insert (c)

Full size image

Figure 4 highlights the surfaces of interest regarding the alignment and specifies in the reported tables how these geometries can be used to set the workpiece coordinate reference system appropriately.

Fig. 4
figure 4

Schematic representation of inserts in which surfaces are highlighted in different colors corresponding to the geometrical features used for alignment; the tables on the right explain how the workpiece reference coordinate system can be generated by combining such geometrical features

Full size image

2.3 Production and assembly

Figure 5a shows an example of plate made of stainless steel, produced through high-precision machining. In this figure, a total of 12 inserts of different types are shown mounted inside the pockets. Figure 5b, on the other hand, displays three real inserts (one for each type), always produced by high-precision machining. As discussed in Section 2.2, the plate design can be modified in terms of size and position of pockets and inserts to adapt to specific needs. In this work, the number of inserts and their sizes have been chosen to utilize the plate for in-depth studies about process dynamics and for process parameter optimization. For the latter use, for example, 12 samples can be fabricated by varying parameters such as laser power, speed, and scanning strategy using design-for-experiment approaches to identify the best combination. The dimensions of the inserts were instead chosen for enabling high-resolution CT scans, as explained in the preceding sections (in particular, Sections 2.1 and 2.2).

Fig. 5
figure 5

Plate with assembled inserts (a) and three inserts with different diameters of the upper part (b); the plate and the inserts are made of stainless steel and produced by high-precision turning and milling

Full size image

The assembly of inserts is schematized in Fig. 6: Figure 6a shows as an example an 8 mm insert already mounted in the plate with the building plane colored in green; Figure 6b shows a cross section of the assembly and Fig. 6c a zoomed detail of it. The insert is kept in position using an M4 screw that is fixed from below. The screw, once fixed, must not exceed the lower plate surface. Such surface, in fact, has to be coupled with its counterpart anchored to the moving piston and any obstacle hindering the coupling must be avoided. The alignment accuracy and repeatability can be achieved by registering the two cylindrical hole portions (one on the insert and one on the pocket) and inserting in the generated hole a precision pin to lock the rotation while screwing the M4 from below.

Fig. 6
figure 6

Schematic representation of the assembly of an insert into the dedicated plate pocket using an M4 screw

Full size image

3 Examples of plate uses for in-process monitoring and investigation of process dynamics

In this section, the possible uses of the plate will be demonstrated with some examples. Section 3.1 shows how the plate can be helpful for in-process PBF-LB/M monitoring, focusing on the perspective correction for off-axis systems and on the enhanced comparison between in-process monitoring and CT data. Section 3.2 presents a method based on the use of the plate and X-ray CT for the accurate evaluation of the so-called “steady state”.

3.1 In-process monitoring and comparison with post-process CT measurements

In Fig. 7a, the plate assembled with 12 inserts is shown positioned above the moving piston before the process start. In Fig. 7b, the plate can be seen in a lowered position, with the inserts surrounded by powder and their top surface aligned with the laser focusing plane.

Fig. 7
figure 7

Plate with assembled inserts mounted in the AM machine (a) and plate covered by steel powder before starting the PBF-LB/M process (b)

Full size image

For enabling accurate process monitoring using off-axis optical systems, the following steps should be applied before starting the process: (i) lower the plate so that the surface containing the holes for image perspective correction aligns with the laser focusing plane, (ii) acquire an image of the plate or a specific area of interest using the selected monitoring instrumentation, employing the same setup that will be used during the process, and (iii) further lower the plate to the level shown in Fig. 7b plus an additional lowering step equal to the nominal layer size. This last step is necessary to deposit the first layer of powder that will be processed by the laser. The above-mentioned steps will be addressed more in detail in the following.

As introduced in Sections 2.1 and 2.2, when dealing with in-process monitoring systems based on off-axis positioning of optical sensors (e.g., visible or near-infrared cameras), the gathered images are affected by perspective deformations. This is visible, for example, in Fig. 8a, which represents an image of the plate acquired before starting the process. This image can be used to compute the perspective transformation parameters, which can be derived by comparing the measured relative distances between the plate’s cylindrical markers to the reference values. In particular, the contour of such markers must be determined and subjected to ellipse least-squares fitting to identify their center coordinates. In this work, the contour determination was carried out by means of the active contour without edges method, by Chan and Vese [28]. The outcomes obtained on the central region of the plate are shown as an example in the inset of Fig. 8a (i.e., red contours and central points). The markers’ coordinates are then used to compute the Euclidean distances between couples of centers, which are then correlated to the corresponding reference values. Once the corrective parameters are determined, they can be stored in a transformation matrix that links the acquired images’ coordinate system to the physical coordinate system of the plate itself, allowing the perspective correction for all the layer-by-layer images acquired thereafter during the process.

Fig. 8
figure 8

Plate image before perspective correction (a) and plate image after correcting the perspective distortions (b)

Full size image

The perspective correction can be applied considering the entire plate, as shown in Fig. 8a, b, where the presence of many reference markers plays a crucial role in achieving accurate results.

On the other hand, since spatial resolution is normally achieved at the expense of the field of view (FOV), it might happen that high-resolution in-process monitoring sensors are limited to small FOVs. To this aim, the integration of markers across the entire building plate enables local correction of perspective distortions, thus allowing the use of high-resolution PBF-LB/M in-process optical monitoring systems composed either by a single sensor focused on a specific plate area or multiple sensors focusing on different plate areas. Figure 9a shows an example of limited FOV with only three inserts monitored by the employed high-resolution imaging device, while the markers used for the local correction of perspective deformations are reported in Fig. 9b with the segmented contours depicted in red. As explained before, the transformation matrix can be computed by exploiting the mathematical relation between the center-to-center Euclidean distances and the physical reference distances. Figure 9c shows the outcome of the local correction applied to the image acquired right before the spreading of the first layer of powder, whereas a close-up view of one of the inserts is represented in Fig. 9d. Figure 9e, f shows a long-exposure image gathered while monitoring the sample fabricated on the same insert of Fig. 9d before and after applying the perspective correction, respectively. In-process detailed observations are therefore enabled by the integration of the markers.

Fig. 9
figure 9

Limited field of view of the powder bed showing three inserts arising from the powder (a), Chan-Vese segmentation carried out on the plate’s markers (b), outcome of the local perspective correction right before starting the process (c), close-up view of one of the inserts (d), and long-exposure monitoring images of the sample fabricated on the same insert before (e) and after (f) the perspective correction

Full size image

In-process monitoring systems can also take advantage of the machined features characterizing the inserts. In fact, a critical aspect when comparing post-process CT results with datasets gathered in-process is the accuracy of the alignment between actual defects and anomalies. In this regard, a previous work [16] pointed out the need for considering part deformations when comparing in-process PBF-LB/M monitoring and post-process CT data. Such work presented an approach based on modeling part deformations using samples having the geometry shown in Fig. 10a, leading to an enhanced comparison accuracy and providing a solution to improve in- and post-process correlations. In such work, a conventional building plate was used and the colored features in Fig. 10a were used to define the sample reference system. However, those features were part of the sample design itself and could consequently be affected by local deformations induced by the AM process. Furthermore, they are characterized by the typically complex PBF-LB/M surface texture, as can be appreciated in Fig. 10b. In this regard, the possibility of using high-precision machined features as those of the inserts allows the definition of a more robust reference system for the fabricated samples, as previously defined in Section 2.2 and in Fig. 4. Figure 10c shows how the designed geometry can now be built directly on the insert. The better definition of the sample reference system allows the further improvement of the registration between in-process monitoring data and CT measurements and the development of more accurate models describing the deformation of the parts.

Fig. 10
figure 10

Sample from [16]: CAD model with colored features originally used for defining the reference system (a); example of CT reconstructed volume (b); CAD model on top of the new plate 12 mm insert, with colored features used for defining the sample reference system (c)

Full size image

3.2 Steady-state investigation in PBF-LB/M using CT

In some cases, it may be necessary to fabricate the desired parts directly on the upper surfaces of the inserts without using intermediate support structures. In such scenarios, it is crucial to consider that the laser powder bed fusion process requires a certain number of layers to reach the so-called “steady state”, defined as the state in which the effective thickness of each individual consolidated layer stabilizes around the nominal thickness value [29]. The steady state can depend on various factors, primarily the type of material and applied process parameters.

The multi-step spiral samples shown in Fig. 11a were used to evaluate the steady state of H13 tool steel, processed with the parameters listed in Table 1. These samples are constituted of different surface portions, each characterized by a different height relative to the building plane (i.e., top surface of the insert). Specifically, heights ranging from 0.02 mm (equal to the layer thickness) to 2 mm were considered, with intermediate steps of 0.02 mm. The height of each portion with respect to the insert top plane was evaluated through CT scanning, using the metrological CT system described in Section 2.2 (Nikon Metrology MCT225) and the acquisition parameters reported in Table 2. An example of CT reconstruction is visible in Fig. 11b. Figure 11c presents the deviation values between the measured height (H) and the nominal height. The accurate evaluation of such deviations was enabled by the possibility of using the machined top plane of the insert as a robust reference, which would be not available if using a conventional plate (in this case, as-built AM surfaces—which are typically irregular and characterized by high roughness—are commonly used as reference to evaluate the steady state [29]). It can be observed that the deviation reaches a plateau oscillating between −0.21 and −0.31 mm after 60 layers, i.e., after 1.2 mm. Consequently, the total height of the produced samples should be compensated for the registered deviation and the features of interest should be positioned above this height of 1.2 mm, which value will however need to be assessed on a case-by-case basis. The large range of oscillation in the steady state is due to the process instability using the selected parameters, which is further demonstrated by the zoomed cross section reported in Fig. 11d. Here, internal and large lack-of-fusion voids were detected by CT under sphere-like formations, which were identified as undesired process by-products such as spatter or balling-related particles, both typically caused by process instability [30, 31].

Fig. 11
figure 11

Multi-step spiral samples produced on 12 mm inserts to assess the steady state (a), CT reconstruction of one of such samples (b), evaluation of the steady state from the deviation between measured and nominal heights of the different surface portions (c), and zoomed view of a CT cross section showing the presence of large sub-surface lack-of-fusion voids (d)

Full size image
Table 1 PBF-LB/M process parameters
Full size table
Table 2 CT scanning parameters
Full size table

4 Conclusions

This paper presented a new concept of building plate for metal laser powder bed fusion, characterized by the following features and advantages:

  • Adequate reference markers are integrated across the entire plate’s base to enable direct and accurate perspective corrections when implementing off-axis optical in-process monitoring systems.

  • The presence of dismountable inserts featuring reference high-precision machined geometrical elements enables the accurate alignment between high-resolution CT reconstructions (down to 3 μm of voxel size, depending on the size of the used insert), in-process gathered data, and building volume characteristics (such as building, coating, and gas flow directions).

  • The achieved accurate alignment can be exploited as a sound basis to improve the comparison between CT dimensional measurements and in-process monitoring data, which is fundamental to understanding if the gathered in-process monitoring signals are correlated to actual defects.

  • The plate can improve CT measurements in tasks where reliable reference geometries are crucial for the measurement accuracy. The example of CT-based measurement of “steady state” in PBF-LB/M was addressed in the paper.

  • The inserts are assembled into a fixed base, so they can be restored after fabrication and CT scans, while the base and the markers for the perspective correction remain unaltered and immediately ready for the subsequent build job. This aspect is relevant for the production rate but also for material and energy savings.

The potentialities of the new building plate concept were demonstrated through examples related to in-process monitoring and post-process X-ray CT measurements, using a specific plate design developed by addressing the identified requirements and adopting a design-for-metrology approach. Nevertheless, even if the design shown in this paper was specific for the PBF-LB/M machine and for the X-ray CT system used in this work, it can be easily adapted to different conditions.

Future publications will be focused on demonstrating other uses of the plate, for example, for developing new and more accurate measurement procedures for quantifying different process- and product-related aspects, useful to increase and deepen the knowledge of the PBF-LB/M processes.